Scientists everywhere use linear regression because that’s what they know, but more often than not the phenomena they study are nonlinear, and a multilayer perceptron can model them. Linear models are blind to phase transitions; neural networks soak them up like a sponge.
